There is some difference between a negative split during a tactic competition (as most part of times happens during top Championships), many times due to a very fast last lap possible because before the race was very slow, and a negative split during a WR or, in any case, one of the best chronometric performances all time.
The important fact is that the most part of WR are done using negative splits, with a progressive acceleration in the second part :
2000m (El Guerrouj, 4:44.79) : 2:23.92 + 2:20.87
5000m (Bekele, 12:37.35) : 7:37.34 (average 2:32.45) the first 3 km + 4:59.99 the last 2 km (2:40.59 + 2:29.42 with the last lap in 57.9)
10000m (Bekele, 26:17.53) : 13:09.42 + 13:08.11 (with last lap in 57.1)
3000m SC (Shaheen, 7:53.63) : 2:36.13 + 2:41.96 + 2:35.54
Half Marathon (Tadese, 58:23) : the last 10 km in 27:08
3000m W (Wang Jungxia, 8:06.11) : 2:42.0 + 2:47.7 + 2:36.4
10000m W (Wang Jungxia, 29:31.78) : 15:05.7 + 14:26.08 (the WR of 5000m became better than the seconf half in 2004 only !) with the last 3 km in 8:17.38 !
About Abdi Bile, I fully agree with Ventolin. May be not exactly 3'26", but 100% under 3'28".
People habe to think that, many times, top athletes didn't have too many opportunities in their life for running, when in top shape, in the best conditions for trying a very important time.
Abdi Bile in that period never had the opportunity of races with rabbits, as later happened to El Guerrouj.
I remember in Oslo, when Andy Norman organised the "Golden Mile", in 1989. The competition had a specific budget, and the theme was the duel between Abdi and Steve Cram.
Said Aouita was in good shape (the next month beated the WR of 3000m becoming the first athlete under 7:30, with 7:29.45 in Koln on 20th August) and wanted to try the WR of Mile. I was present at a very animated discussion between Said and Andy, because Norman told him "You can be the best in the World, you can chose the distance you want to run, may be 800 or 1500 if you want or 3000 or 5000, BUT NOT THE MILE. This MUST be the return match between the World Champion of 1500 and the WR holder of mile, and NOBODY else can go in the middle !".
These were the rules of the circuit, different from today. In that competition, Abdi won in 3:49.90 , Wilfred Kirochi was second in 3:50.49, Joe Falcon was 3rd in 3:51.07 and Cram was only 4th in 3:51.58 (and Aouita won 3000m in 7:34.79).
I want also to explain that it's not possible to compare the value of performances done in different periods.
For example, it's true that Abdi Bile had only a PB of 3:30.55, but the WR was only 3:29.46, and the organisation of pacers was according to that level.
Abdi lost OG 1988, that he could win being absolutely stronger than Peter Rono, for a stress fracture in his left foot. Later, he had a lot of problems with his family in Somalia, losing the best opportunities of continuity in training and in comopetitions. However, was still able to have a good come back in 1993, when won the bronze medal in World Championships.
Speaking about competitions in normal meeting, I remember El Guerrouj winning in Torino (2003, the week before his debut in 5000m in Ostrava) a race of 3000m in 7:30.23 (6th of June). He asked for a rabbit able running the first 1500m in 3:40.0, having the idea to try the WR. During warm up, the second pacer had an injury, and the competition went with the first rabbit only, very week. So, the first 1500m were run in 3:52.6. Hicham was very upset, and started to run faster changing speed quickly. His second 1500m were in 3'37"6, completely alone !
The next week, in Ostrava, it was the first time Stephen Cherono (not yet Shaheen) ran a serious 5000m. I told him that, in my opinion, Hicham wanted to try to beat the WR, looking at the shape showed in Torino (he asked for a pace of 7'37" the first 3 km). So, I explained Stephen "Hicham is use to run against time, not against other competitors, and this is the reason he lost in Sydney from Ngeny. So, when the pacers finish their work after 3000m, he thinks to be alone. Instead, you must be with him, and, if you are able to stay during the 4th km, he starts to become nervous, reducing the pace. In this case, you must show him you are fresh, going on his side with two laps to go, and after you go behind him again, waiting the last furlong".
Stephen did exactly this, and won in 12'48"81 against 12'50"24, that after that race never they bettered.
At the end, only one question for Ventolin :
If you accept the idea that, for a top athlete is possible to improve of 3.0 in 1500m in one season, why do you think that, when an athlete NEW can improve, TODAY, is necessary to use doping ? In your opinion, there is more doping today (when athletes are no more able to run fast, in 1500m, like 10-20 years ago) than in the past ?
So, in your opinion, the athletes of the past WERE STRONGER THAN THE ATHLETES OR TODAY, or were better doped, or there was a period when the goal was to try everytime to run fast, with a top organisation of the pace in every competition ?
Because the fact is that, during the last 5 years, the only athlete going in top 10 all-time in 1500m is Silas Kiplagat last year, and the explanation is in one of the reasons I wrote before.